分布式数据库系统的体系结构与设计（PPT讲稿）Architecture and Design of Distributed Database Systems

Architecture and design of Distributed database Systems WAMDM Cloud Computing Group Haiping Wang 2010-05-08 2021/2/1 WAMDM Cloud Group

Architecture and Design of Distributed Database Systems WAMDM Cloud Computing Group Haiping Wang 2010-05-08 2021/2/11 WAMDM Cloud Group 1

Outline Architecture Client/Server, P2P DDBMS MDBS Design Strategies Issues Fragmentation A| ocatⅰon Conclusion 2021/2/1 WAMDM Cloud Group

Outline • Architecture – Client/Server, P2P – DDBMS,MDBS • Design – Strategies – Issues – Fragmentation – Allocation • Conclusion 2021/2/11 WAMDM Cloud Group 2

Client/Server Reference Architecture User Application Begin at 1990s aELInterface Program Communication at the level of sQl Client DBMS statement Communication Software Client SQL Result Application, user interface, DBMS Client queries relation responsible for cache management( cached data and cached transaction locks) Communication software Possible for consistency checking of user p Semantic Data Controller queries(not common Query Optimizer · Server Transaction Manager Query processing, optimization Recovery Manager transaction and storage management g L Runtime Support Processor One server system Comparison with centralized DB Multiple server client manages its own connection(heavy client Client only know it "home server"(light client Database 2021/2/1 WAMDM Cloud Group

Client/Server Reference Architecture • Begin at 1990’s • Communication at the level of SQL statement • Client – Application , user interface, DBMS Client responsible for cache management( cached data and cached transaction locks) – Possible for consistency checking of user queries(not common) • Server – Query processing, optimization, transaction and storage management – One server • Comparison with centralized DB – Multiple Server • client manages its own connection(heavy client) • Client only know it “home server”(light client) 2021/2/11 WAMDM Cloud Group 3

Physical data Organization in P2P Distributed Systems Data usually fragmented and replicated Three layer organization ES LiS (local internal schema) GCS LCS(local conceptual schema gCs( global conceptual LCS schema ESs external schema), for user LIS applications and access 2021/2/1 WAMDM Cloud Group

Physical Data Organization in P2P Distributed Systems • Data usually fragmented and replicated • Three layer organization – LIS (local internal schema) – LCS(local conceptual schema) – GCS ( global conceptual schema) – ESs (external schema), for user applications and access 2021/2/11 WAMDM Cloud Group 4

Functional Schematic of an Integrated DDBMS Administrator Global Administrator database schema Administrator For data independence database schema >Extended the anSi/sparc processo GD/D for global mapping Local Data Management Locai Data Management >LD/D for local mapping Database Administrator processor-1 processor-1 Administrator schema processor processor Int storage/ intemal transtormation transformation Mem cloud Group

Functional Schematic of an Integrated DDBMS ➢For data independence ➢Extended the ANSI/SPARC ➢GD/D for global mapping ➢LD/D for local mapping 2021/2/11 WAMDM Cloud Group 5

Components of a ddbms User processor User interface handler requests Semantic data controller ROCESSOR ser Interfac Extem Global query optimizer and decomposer Overy Distributed execution monitor Global Execute Data processor OCESSOR Local query optimizer Query Processc Local recovery manager Recovery Manag Log Run-time support processor Local Inter Processor 2021/2/1 WAMDM Cloud Group

Components of a DDBMS • User processor – User interface handler – Semantic data controller – Global query optimizer and decomposer – Distributed execution monitor • Data processor – Local query optimizer – Local recovery manager – Run-time support processor • 2021/2/11 WAMDM Cloud Group 6

MDBS Architecture With a gcs Without gcs GES ges IGEs E ES n LES, LES, LES Multidatabase Local system LCS ayer LCS LCS LCS LIS, LIS LIS >GCS, GES may use different data model and language Local system layer >Homogeneous >Unilingual(example: MULITIBASE Multidatabase layer Multilingual(permit each user to access the global data by means of external schema) >heterogeneous WAMDM Cloud Group

MDBS Architecture With a GCS Without GCS 2021/2/11 WAMDM Cloud Group 7 ➢GCS,GES may use different data model and language ➢Homogeneous: ➢Unilingual(example:MULITIBASE) ➢ Multilingual(permit each user to access the global data by means of external schema) ➢heterogeneous ➢Local system layer ➢Multidatabase layer

Difference between mdbs and ddbms Definition difference DDBMS MDBS Global conceptual schema The conceptual view of the The collection of some of (fundamental difference) entire database, mapping the local db that wants to the global schema to local share, mapping the local conceptual schema onceptual schema to a global schema Global database Equal to the union of local A subset of the union of databases the local databases Design strategy Usually designed by top- Usually designed by down process bottom-up process 2021/2/1 WAMDM Cloud Group

Difference between MDBS and DDBMS Definition difference DDBMS MDBs Global conceptual schema (fundamental difference) The conceptual view of the entire database, mapping the global schema to local conceptual schema The collection of some of the local DB that wants to share, mapping the local conceptual schema to a global schema Global database Equal to the union of local databases A subset of the union of the local databases Design strategy Usually designed by top￾down process Usually designed by bottom-up process 2021/2/11 WAMDM Cloud Group 8

Outline Architecture Client/Server, P2P DDBMS. MDBS Design Strategies Issues Fragmentation Allocation Conclusion 2021/2/1 WAMDM Cloud Group

Outline • Architecture – Client/Server, P2P – DDBMS,MDBS • Design – Strategies – Issues – Fragmentation – Allocation • Conclusion 2021/2/11 WAMDM Cloud Group 9

Framework of distribution evel of sharing Share nothing Access pattern Share data D Share da re datat progra Static Partial ° Access pattern nformation Stati ata Level of knowledge Dynamic vel of knowledge or Complete Data nformation program access pattern Know nothing The designer know partial information The designer know complete information 2021/2/1 WAMDM Cloud Group

Framework of Distribution • Level of sharing – Share nothing – Share data – Share data+ program • Access pattern – Static – Dynamic • Level of knowledge on access pattern – Know nothing – The designer know partial information – The designer know complete information 2021/2/11 WAMDM Cloud Group 10